Author(s): Sebastian Zimmeck, Ziqi Wang, Lieyong Zou, Roger Iyengar, Bin Liu, Florian Shaub, Shomir Wilson, Norman Sadeh, Steven M. Bellovin, Joel Reidenberg

Download: Paper (PDF)

Date: 27 Feb 2017

Document Type: Reports

Additional Documents: Slides Video

Associated Event: NDSS Symposium 2017


Mobile apps have to satisfy various privacy requirements. Notably, app publishers are often obligated to provide a privacy policy and notify users of their apps    privacy practices. But how can a user tell whether an app behaves as its policy promises? In this study we introduce a scalable system to help analyze and predict Android apps    compliance with privacy requirements. We discuss how we customized our system in a collaboration with the California Office of the Attorney General. Beyond its use by regulators and activists our system is also meant to assist app publishers and app store owners in their internal assessments of privacy requirement compliance.

Our analysis of 17,991 free Android apps shows the viability of combining machine learning-based privacy policy analysis with static code analysis of apps. Results suggest that 71% of apps that lack a privacy policy should have one. Also, for 9,050 apps that have a policy, we find many instances of potential inconsistencies between what the app policy seems to state and what the code of the app appears to do. In particular, as many as 41% of these apps could be collecting location information and 17% could be sharing such with third parties without disclosing so in their policies. Overall, each app exhibits a mean of 1.83 potential privacy requirement inconsistencies.